Ask Northwell Health CEO Michael Dowling what worries him about artificial intelligence and he'll share his list of concerns about the speed of development and potential to perpetuate biases.
Dowling said he believes AI will create new therapies, prediction tools and diagnostics. Still, he said, it’s important for everyone in healthcare to slow down because there is a lot of potential to do harm.
Related: Why health systems are buying into genomics
“Progress is good, we’re all a little impatient. We all want to do more all the time quickly,” Dowling said. “But progress too fast can be dangerous because there’s a lot we don’t know. There’s a lot we don’t understand."
Dowling gave a keynote address Wednesday at an annual summit hosted by regional trade group Digital Health New York and healthcare investor AlleyCorp. The event attracted around 350 digital health company founders, health system leaders, investors and others.
New Hyde Park, New York-based Northwell Health hasn't exactly sat on the sidelines with AI. Dowling touted its partnership with startup creator Aegis Ventures. Aegis, which forms, invests and operates digital health companies, launched a digital consortium with Northwell, Houston-based Memorial Hermann Health System and seven other health systems in April to create AI-focused companies that will address clinical pain points. Northwell and Aegis also launched healthcare AI platform company Ascertain and retinal AI imaging company Optain.
Dowling said the system has been inundated with pitches for AI-focused solutions and called it overwhelming. AI companies need to “calm down” and ensure their models aren’t making things worse for providers, he said. He also expressed concerns with how AI could perpetuate biases against people of diverse backgrounds.
“When we apply the standards of research based on white males to women and minorities, it doesn't necessarily automatically apply. If you use all these datasets to create an algorithm, you will perpetuate that which you don't want to perpetuate,” Dowling said. “That's why you're taking time matters.”
Dr. Brendan Carr, CEO of New York City-based Mount Sinai Health System, told attendees the health system has spun off multiple AI companies including PreciseDx, an oncology company that received $20.7 million in a Series B funding round in August. It also has created several in-house algorithms including predictive models to identify patients at risk for clinical deterioration.
He called on digital health entrepreneurs to work with health systems to create a refined process to develop and deploy AI models, particularly when it could affect clinical care.
“We at academic health systems, if we're going to try and take novel things, implement them, measure the impact of them, and then scale then, then we have to do it in a much more organized way,” Carr said. “It’s on us to build better processes and it’s a little bit on you [digital health entrepreneurs].”
Carr and others noted the challenges associated with implementing AI within health systems.
“I can think of multiple times where we’ve had an enterprise-wide implementation and everybody's well intentioned, but true integration of the workflow came down to the font sizes and the colors,” said Dr. Peter Fleischut, chief information and transformation officer at New York-Presbyterian.
AI vendors at the event acknowledged the difficult environment they encounter when deploying their technology at health systems. Dr. Shiv Rao, CEO of Generative AI clinical documentation company Abridge, said it was important to understand that one health system can be very different from another. Abridge received $150 million in funding in February.
“You have to unpack what a health system looks like,” Rao said. “They might have a rural community health hospital. They might have a large academic in the city. They might have a small provider that they bought recently. It looks very heterogeneous. Their patients might speak multiple languages. You have to serve across multiple settings, across all the specialties that exist in a large system, across all the different spoken languages.”
